# Dump list for the rd_cull_ab ctest's MSAA runs (port/render/test/rd_cull_ab.cmake,
# -DMSAA_LIST): each dump here is compared byte for byte with every scene draw replayed and
# with the visibility test, at Enhanced 4:3 2x under --aa msaa4, beside the first dump of the
# corpus.  Chosen where the MSAA epoch closes inside a list by a read of SCENE (a texture read
# or a DATE snapshot, rd_replay.c rd__transition), not by the shadow count's pass at a list's
# start: the visibility test must keep a draw that could close the epoch (rd_vis.c
# rd__vis_msaa_may_close; docs/port/RENDER.md "Culling", "Under MSAA").  Over the 2026-11-04
# corpus (tools/occlusion_corpus.txt, tools/native_frame_dumps.txt and the v0.10.1 evidence)
# no dump closes the epoch inside lists 0 to 2: every gameplay frame closes at list 3 (the
# shadow count's pass, or a read there), the phone dumps of the GL cull fix at list 12 or never.
# MACHINE-LOCAL: paths on the maintainer's machine; the dumps are never committed.  Missing
# files are reported and skipped.  Relative paths are relative to this file; '*' globs are
# allowed.
#
# issue 182's frame (the MSAA report): the epoch closes at list 3 by a read of SCENE
/primary/dev/ico-pc/build-host/tmp/v0101/evidence/182/frame-20261103-163115-v3225.rddump
# issue 102's frame (v0.7.9): the epoch closes at list 3 by a read of SCENE
/primary/dev/ico-pc/build-host/tmp/evidence-v079/102/frame-20261015-161933-v4251.rddump
